Skip to content

DEV Community

Mark Huang

404 bio not found

Joined on Jun 2, 2026 https://markhuang.ai

Mark Huang

Sep 10

DeepSeek's Pro Endpoint Will Stop Meaning Pro

6 min read

Want to connect with Mark Huang?

Create an account to connect with Mark Huang. You can also sign in below to proceed if you already have an account.

Already have an account? Sign in

Mark Huang

Sep 5

GPT-6 Astra Reprices the Whole Request After 272K Tokens

5 min read

Mark Huang

Sep 4

Astra's ARC Score Moves 45.1 Points When the Harness Changes

6 min read

Mark Huang

Sep 2

Fable 5.1 Rebuilt Union Square and Left Its Misses in Git

5 min read

Mark Huang

Sep 2

Muse Spark 1.3 Charges 21 More for Output It Won't Train On

5 min read

Mark Huang

Sep 2

Gemini 3.7 Flash Called 12% of Poisonous Mushrooms Edible

5 min read

Mark Huang

Sep 2

Muse Spark 1.3 Uses 25% Fewer Tokens. Its Max Mode Still Has to Wait.

6 min read

Mark Huang

Sep 2

Gemini 3.8 Flash and Cyber Share a Core, Then Split at the Guardrails

5 min read

Mark Huang

Sep 2

Gemini 3.8 Flash Kept 3.7's Price. Higher Effort Still Runs the Meter.

5 min read

Mark Huang

Sep 2

Nori A3 Is $1,688. Who Teaches It to Fold the Laundry?

6 min read

Mark Huang

Sep 2

Claude Fable 5.1 Made Thinking History Append-Only for New API Accounts

6 min read

Mark Huang

Sep 1

Haiku Beta 6 Nearly Halved a Five-Hour Build. I'm Still Using a Spare Disk.

5 min read

Mark Huang

Sep 1

Hy4 Defaults to Deep Reasoning. Tencent Says It Overthinks.

5 min read

Mark Huang

Aug 29

GLM-5.3 Waited Two Weeks. Now the Weights Can't Be Recalled.

6 min read

Mark Huang

Aug 29

Claude's Watermark Can Flag Claude, Not Settle Authorship

6 min read

Mark Huang

Aug 17

Claude's System Prompt Is Public. Why Can't It Explain Claude Code?

5 min read

Mark Huang

Aug 15

Opus 5 Finishes the Task, Then Changes the Brief

5 min read

Mark Huang

Aug 14

At 750 Tokens a Second, GPT-5.6 Sol Moves the Bottleneck

5 min read

Mark Huang

Aug 14

Gemini 3.7 Flash Is Half Price Until 2027

5 min read

Mark Huang

Aug 13

LLMs Can Search More Proofs. Can They Choose the Right One?

6 min read

Mark Huang

Aug 13

Facebook Ads Turned an Open-Source Filter Into a Staffing Problem

5 min read

Mark Huang

Aug 13

Grok 4.6 Took 53 Turns. Claude Opus 5 Took 103.

6 min read

Mark Huang

Aug 11

Claude Code Allegedly Put a Real Email in User-Agent

6 min read

Mark Huang

Aug 8

DeepSeek V4 Flash Solved 61.4% for Four Cents. Now Test the Workflow.

6 min read

Mark Huang

Aug 7

The Agentic Index Ranks 24 Models. I Still Need a Failure Test.

6 min read

Mark Huang

Aug 7

TIME Built an Ad Slot Only AI Bots Can See

6 min read

Mark Huang

Aug 5

The AI Agent Did Not Escape. The Cyber Eval Let It Out.

6 min read

Mark Huang

Aug 5

LLMs Reward Experts. Who Trains the Next Ones?

6 min read

Mark Huang

Aug 3

Go 1.27 Adds Generic Methods. The Upgrade Risk Is in the Defaults

6 min read

Mark Huang

Aug 1

Claude Was Told the Internet Was Fake. Three Companies Were Real.

6 min read

Mark Huang

Aug 1

GPT-5.6 Sol Ran a Business for 24 Hours. It Optimized the Score.

6 min read

Mark Huang

Aug 1

Gemini Robotics 2 Can Make Hundreds of Decisions. I Care About the One That Stops It.

6 min read

Mark Huang

Jul 30

AI Can Read the Literature. It Still Can't Tell Which Papers to Trust.

5 min read

Mark Huang

Jul 24

The Requirements File Was Clean. The Git Hook Was the Trap.

7 min read

Mark Huang

Jul 21

Kimi K3 Got Close. Anthropic's Moat Test Starts After the Benchmark.

6 min read

Mark Huang

Jul 21

Kimi Work Can Run 300 Agents. I Want the Receipts.

6 min read

Mark Huang

Jul 20

American AI Is Losing the Download Race. Is That the Market?

6 min read

Mark Huang

Jul 20

Five Microservices, Three Engineers: Perfection Wasn't the Problem

5 min read

Mark Huang

Jul 20

GPT-5.6 found a WordPress RCE for $25. The human review took longer.

6 min read

Mark Huang

Jul 19

QwenCloud's 40% Discount Comes With Two Quota Clocks

8 min read

Mark Huang

Jul 19

Qwen3.8 Promises Open Weights. Today, the Preview Stays Inside Alibaba.

8 min read

Mark Huang

Jul 19

Codex's Reset Tracker Hit 35. Why Is the Meter Still a Mystery?

7 min read

Mark Huang

Jul 17

Google's Custom Search API Dies in 2027. A Drop-In Isn't a Migration

8 min read

Mark Huang

Jul 17

Kimi K3 Scored 57. The 130 Million Output Tokens Are the Catch.

8 min read

Mark Huang

Jul 14

I Keep Asking Why: The Mindset I Want to Keep in the AI Era

12 min read

Mark Huang

Jul 13

I'm Aggressively Trying to Replace Myself With AI

14 min read

Mark Huang

Jul 13

Claude Turns the Same TypeScript Into 73% More Tokens. That Still Isn't the Bill.

7 min read

Mark Huang

Jul 13

Claude Fable 5 Got Another Week. Super Dario Made the Deadline a Boss Fight

8 min read

Mark Huang

Jul 13

LLMs Are Useful Without the Hype

6 min read

Mark Huang

Jul 13

GPT-5.6 Sol Scores Higher. Why Does My Weekly Limit Vanish Faster?

24 min read

Mark Huang

Jul 12

Terminator 2 Made CGI Earn Every Shot

8 min read

Mark Huang

Jul 12

Model Build-Offs Need Failure Rates, Not Trophies

8 min read

Mark Huang

Jul 12

OpenAI Forked Git. The Empty Diff Is the News

9 min read

Mark Huang

Jul 9

AI Bookkeeping Needs a Harness

8 min read

Mark Huang

Jul 9

Hy3 Makes Price Part of the Eval

8 min read

Mark Huang

Jul 9

Bun's Rust Rewrite Is the Validation Test

8 min read

Mark Huang

Jul 8

AI Heat Needs a Neighbor

8 min read

Mark Huang

Jul 7

98% Support Still Needs a Door

8 min read

Mark Huang

Jul 6

The Token Sticker Price Is a Trap

8 min read

Mark Huang

Jul 6

AI Makes the Average Too Cheap

8 min read

loading...